J Abnorm Child Psychol DOI 10.1007/s10802-015-9999-5

Adolescent-Parent Attachment and Externalizing Behavior: The Mediating Role of Individual and Social Factors Sanne L. A. de Vries & Machteld Hoeve & Geert Jan J. M. Stams & Jessica J. Asscher

# The Author(s) 2015. This article is published with open access at Springerlink.com

Abstract The aim of this study was to test whether the associations between adolescent-parent attachment and externalizing problem behavior of adolescents were mediated by adolescent cognitive distortions, self-esteem, parental monitoring and association with deviant peers. A total of 102 adolescents (71 % male; aged 12–19 years) at risk for developing delinquent behaviors reported on attachment, parental monitoring, aggressive and delinquent behavior and peers. Mediation effects were tested by using structural equation modeling. Different pathways were found depending on the type of externalizing behavior. The association between attachment and direct and indirect aggressive behavior was mediated by cognitive distortions. The relation between attachment and delinquency was mediated by deviant peers and parental monitoring. We argue that clinical practice should focus on the attachment relationship between adolescent and parents in order to positively affect risk and protective factors for adolescents’ aggressive and delinquent behavior.

Keywords Attachment . Externalizing behavior . Self-esteem . Cognitive distortions . Deviant peers . Parental monitoring

Electronic supplementary material The online version of this article (doi:10.1007/s10802-015-9999-5) contains supplementary material, which is available to authorized users. S. L. A. de Vries : M. Hoeve : G. J. J. M. Stams : J. J. Asscher Research Institute of Child Development and Education, University of Amsterdam, Nieuwe Achtergracht 127, 1018 WS Amsterdam, The Netherlands S. L. A. de Vries (*) Research Institute of Child Development and Education, University of Amsterdam, P.O. Box 15780, 1001 NG Amsterdam, The Netherlands e-mail: [email protected]

Meta-analytic studies by Fearon et al. (2010) and Hoeve et al. (2012) have shown that adolescents’ attachment to their parents is associated with concurrent and later aggression and delinquency. Few studies, however, have examined the mechanisms that could explain the association between attachment and these externalizing behaviors. Adolescents have developed mental representations of self and others in attachment relationships with their parents that shape both individual and social functioning. In the present study, we empirically test whether risk and protective factors of individual functioning (i.e., cognitive distortions and self-esteem) and social functioning (affiliations with deviant peers, and parental monitoring through adolescents’ self-disclosure) mediate the association between attachment and externalizing behavior (i.e., aggression and delinquency).

Mediation Through Cognitive Distortions Children internalize both secure and insecure patterns of their relationships with caregivers as mental representations or internal working models of attachment (Dykas and Cassidy 2011; Pietromonaco and Barrett 2000), which influences the way children interact with their environment (Bowlby 1973). Individuals with secure internal working models process a broad range of positive and negative attachment-relevant experiences, and their mental schemas represent a coherent integration and organization of these experiences (Bowlby 1969; Pietromonaco and Barrett 2000). Security of attachment facilitates cognitive abilities (e.g., memory and comprehension) and social understanding (Fonagy and Target 1997). Insecure attachment organizations are characterized by the defensive exclusion of information or inability to integrate different types of information about attachment experiences, which may lead to distorted communications and negative expectations of self in relationship with others (Cassidy et al. 1996;

J Abnorm Child Psychol

Dodge 1993; Shumaker et al. 2009). Children internalize negative experiences with their parents as insecure internal working models of attachment (Blatt and Homann 1992), which have been linked to poor mentalizing abilities (Fonagy and Target 1997), hampering perspective taking and making adolescents vulnerable to egocentric bias and self-serving cognitive distortions, defined as Binaccurate attitudes, thoughts or beliefs concerning own or others’ behavior^ (Gibbs et al. 1995, p. 165). These types of distortions buffer the self from blame or negative self-concept, which reinforces aggression or other forms of antisocial behavior (Barriga et al. 2000). Helmond et al. (2014) showed in their meta-analysis that cognitive distortions are moderately associated with both aggression and delinquency.

Mediation Through Self-Esteem Attachment to parents has been shown to be associated with adolescents’ self-esteem (e.g., Armsden and Greenberg 1987; Gamble and Roberts 2005; Lee and Hankin 2009; Noom et al. 1999; Papini and Roggman 1992; Paterson et al. 1995; Roberts et al. 1996), defined as self-judgments of personal worth and global feelings of competence and self-acceptance (Rosenberg 1965). Through attachment relationships children develop a working model of the self, which consists of generalized perceptions of competence and self-esteem (Greenberg et al. 1993). Children who perceive their parents as being responsive and available are likely to internalize a sense of their own self-worth and expect that others will attend to their needs (see Gerlsma et al. 1996). In contrast, children with insecure working models of attachment view others as untrustworthy or unavailable, which in turn leads to a lack of confidence in self and others (Gamble and Roberts 2005; Gomez and McLaren 2007). The link between low self-esteem and externalizing problems has been well established in empirical research (e.g., Donnellan et al. 2005; Trzesniewski et al. 2006), and may be explained in different ways. According to Rosenberg (1965), low self-esteem weakens ties with society, and low engagement with society in turn decreases conformity to social norms and increases delinquency (Hirschi 1969). It has also been suggested that adolescents with low self-esteem show various forms of antisocial behavior, including aggression, as a way of enhancing their self-worth (Kaplan 1980).

Mediation Through Parental Monitoring Several studies have suggested that attachment is related to self-disclosure (Keelan et al. 1998; Pistole 1993), that is, youths’ tendencies to provide unsolicited information (Kerr and Stattin 2000). Individuals with secure attachments

experience a sense of worthiness, which contributes to more engagement in self-disclosure (Keelan et al. 1998). Selfdisclosure involves a significant amount of trust. Trust is related to the understanding one has of others’ likely responses to personal vulnerability, also referred to as an internal working model of relationships with others (Mount 2005). In particular, trust in others has been linked to the amount of information self-disclosed to another (Levin and Gergen 1969; Pearce 1974). A positive and trusting relationship between parents and adolescents creates an open way of communication about adolescents’ daily activities, thoughts and feelings (Deković et al. 2004). Kerr and Stattin (2000) found that adolescent disclosure contributes to greater parental knowledge of adolescents’ whereabouts. In contrast, insecure representations of attachment with parents could lead to less self-disclosure and parental monitoring (Branstetter et al. 2009). Attachment insecurity has been found to predict greater reluctance of adolescents to provide their parents information on their whereabouts and activities (Kerns et al. 2001; Sampson and Laub 1994). Insecurely attached adolescents tend to spend less leisure time in parental company and are more attracted by unsupervised peer settings (Kerr and Stattin 2000). In addition, insecurity may lead to externalizing behavior by causing hostility toward parents or efforts to minimize conscious attention directed toward parents, either of which may reduce behavioral parental control (Allen et al. 1998). Several studies have concluded that low levels of parental monitoring and lack of knowledge are associated with adolescents’ involvement in a range of antisocial and delinquent behaviors (see Crouter and Head 2002; Dishion and McMahon 1998; Patterson 1986). Inconsistent and erratic supervision by parents promotes deviant attitudes and behaviors in their children (Akers 2000).

Mediation Through Deviant Peers Empirical support exists for the association between attachment and peer affiliations (Benson et al. 2006; Warr 1993). Representations of relationships with parents shape a child’s core strategy of regulating his/her emotions, thoughts and behaviors in close relationships, such as friendships (Bowlby 1973). According to Hirschi’s (1969) social control theory adolescents who are strongly attached to their parents may be less influenced by deviant peers. These adolescents are more prone to seek out nondelinquent peers to avoid parental disapproval or because their parents actively regulate their children’s friendships to avoid undesirable peers (Warr 1993). Youth with insecure attachment relationships have poorer levels of social competence and more negative friendships (Shulman et al. 1994). Negative interactions with parents interfere with effective functioning of a secure base from which adolescents can form friendships, which hampers adolescents’

J Abnorm Child Psychol

ability to establish positive friendships (Shomaker and Furman 2009). Moreover, parental rejection or absence of close bonds with parents leads to an adolescent’s rejection of commitment to conventional values. Subsequently, adolescents rejecting conventional values are more likely to associate with peers who support unconventional standards. In turn, these peers act as role models in learning or reinforcing delinquent behavior that adolescents tend to imitate (Akers 2000). Many studies have considered that affiliations with antisocial and deviant peers are related to various problematic outcomes during adolescence, such as high levels of aggression (Benson and Buehler 2012; Capaldi et al. 2001), police arrests (Patterson et al. 2000), and other forms of antisocial behavior (Ardelt and Day 2002; Reitz et al. 2006; Vitaro et al. 2000).

Etiology of Aggressive and Delinquent Behavior Currently, several studies and classification systems for child and adolescent psychopathology distinguish between aggressive and delinquent behavior, because these two forms of externalizing behavior seem to differ in several aspects. Firstly, several studies showed that aggressive and delinquent behavior are distinct at the etiologic level. Although the interplay of genetics and the environment influences both types of antisocial behavior (aggression and delinquency), genetic influences were suggested to be greater for aggressive antisocial behavior than for nonaggressive antisocial behavior (Eley et al. 2003). Moreover, Tackett et al. (2005) found that shared environmental influences play a significant role in rule-breaking behaviors (delinquency). Other studies have also found a substantial genetic component (around 65 %), but no significant shared environmental component for aggression, whereas delinquent behavior has shown a moderate genetic component (around 35 %) and a moderate shared environmental component (around 35 %; e.g., Edelbrock et al. 1995; Eley et al. 2003). Further, aggressive behavior has been shown to be more stable over time compared to delinquent behavior: after about age 10, aggressive behavior declines, whereas delinquent behavior increases until about age 17 (Stanger et al. 1997). Additionally, aggressive adjudicated youth showed greater deficits in executive neuropsychological functions (such as reasoning, problem solving and planning) than nonaggressive adjudicated youth (Moffitt and Henry 1989). Aggressive behavior could also be divided in several subtypes on the basis of different developmental trajectories, antecedents, and consequences. First, aggression incorporates not only the infliction of physical harm, but also consists of more subtle forms of aggressive behavior, such as social exclusion. These subtle forms of aggression are referred to as indirect aggression, relational aggression, and social aggression (Vitaro et al. 2006). The different developmental trajectories of different types of aggressive behavior are

demonstrated by the overtness-covertness dimension of antisocial behavior (Loeber and Schmaling 1985). The covert pathway (indirect aggression) consists of hostility, irritability, suspicion and anger, whereas the overt pathway (direct aggression) consists of verbal or physical aggression, such as fighting (Lange et al. 1995).

The Present Study In summary, previous research has shown that externalizing behavior of adolescents can be explained by the presence of cognitive distortions, low levels of self-esteem, low degree of parental monitoring and affiliations with deviant peers, which in turn can be explained by poor attachment quality. However, to our knowledge, previous studies have not examined whether the association between attachment and externalizing behavior is mediated by any of these factors (simultaneously). Only Simons et al. (2001) and Gomez and McLaren (2007) found that levels of self-esteem mediated the relation between attachment and aggressive behavior of adolescents. In the present study, we will not only examine mediating effects of individual mechanisms (self-esteem and cognitive distortions), but also of social mechanisms (i.e., affiliation with peers and parental monitoring). Given that there are distinct patterns of antisocial behavior (multidimensional construct), we differentiate between direct and indirect aggression and delinquent behavior. Based on earlier research (e.g., Tackett et al. 2005), we expect that social mechanisms, considered as ‘environment influences’, play a more important role in the relation between attachment and delinquency, whereas individual mechanisms are expected to be more influential in explaining the relation between attachment and aggressive behavior. Therefore, we will examine two separate mediation models for delinquency and aggression.

Method Participants and Procedure Data were obtained from adolescents who were referred to a youth care organization and enrolled in programs for youth at risk for criminal behavior and youth who had committed minor delinquent acts in the period of 2011–2013. Participation in these programs was voluntary. Treatment professionals (specialized in child psychology) determined whether adolescents were eligible for participation on the basis of following criteria: age 12 to 23 years, experiencing problems in multiple life domains (school, family, peers, leisure time), and at risk for the development and progression of a deviant life style. For example, predelinquents with antisocial behavior, first time offenders, and adolescents with mainly minor (first) police contacts and offenses (such as, inflicting damage or

J Abnorm Child Psychol

destroying property on purpose, shoplifting and joyriding) were eligible for participation. Juveniles with a longer history of delinquent acts or showing severe psychopathology before age 12 were excluded from participation. After screening (for eligibility) and consent to participate, adolescents were asked to complete a questionnaire. The Ethics Committee of the University of Amsterdam (2011-CDE-01) approved the study design, procedures and informed consent. A total of 160 adolescents were eligible and approached for participation. Finally, data of the first measurement on demographics, parental attachment and externalizing behavior were available for 102 adolescents (63.8 %). 36.2 % (n=58) of the included adolescents declined to participate on the first measurement because of several reasons (8 parents and 20 juveniles did not consent to participate; 15 juveniles could not be reached; 15 other reasons, such as migration/language problems). Results of independent t-tests and chi-square tests showed no differences between participants and nonparticipants in age, ethnic background and gender. All participants, aged 12 to 19 years, lived in the urban area of Amsterdam (the Netherlands). The sample of participants mainly consisted of the major ethnic groups in Dutch large cities: native Dutch (n=20; 20 %), Moroccan (n=26, 26 %), Turkish (n=9; 9 %), and Surinamese (n=26, 26 %). The remaining participants had other ethnic backgrounds (n=21; 21 %). Ethnic group membership was defined by the birth country of both parents and the adolescent (native Dutch: if both parents were born in the Netherlands). About 34 % of the participants reported living with both parents, 53 % reported living with one parent (mother or father), 3 % reported living partly with both parents, and 10 % reported living with other relatives. Additional characteristics of the total sample are presented in Table 1. Table 1

Sample characteristics for the total sample (N=102)

Age Gender (Male) Cognitive distortions Parental monitoring Self-esteem Deviant peers Direct aggression Indirect aggression Delinquency Trust (attachment) Communication (attachment) Alienation (attachment)

M

SD

15.52 72a 2.61 2.89 3.07 1.64

1.53 70.6b 0.68 0.62 0.62 0.72

0.61 0.45 4.24 3.13 2.78 3.26

0.23 0.22 4.45 0.81 0.85 0.63

Attachment = Trust, Communication and Alienation a

n

b

%

Measures Parental Attachment The attachment relationship between the adolescent and parent was assessed using the short version of the ‘Inventory of Parent and Peer Attachments’ (IPPA; Armsden and Greenberg 1987). This instrument was designed to assess the extent to which adolescents felt secure by measuring the adolescents’ trust in availability and sensitivity of the attachment figure, the quality of communication and the extent of anger and alienation in the relationship with the attachment figure. The IPPA is a 12-item self-report questionnaire using a 4-point Likert-type scale (1 = almost never, to 4 = almost always). Examples of statements for each scale are: BIf my parent knows something is bothering me, he/she asks me^ (communication); BMy parent respects my feelings^ (trust); BI don’t get much attention from my parent^ (alienation). The IPPA proved to be reliable and valid in previous studies (Armsden and Greenberg 1987; Deković and Meeus 1997; Raja et al. 1992). Based on the dissatisfactory outcomes of reliability analyses and low item-total correlations on the subscales of communication (α=0.53) and trust (α=0.32), two items were deleted (communication scale: Bmy parents have their own problems, so I don’t bother them with mine^; trust scale: BI wish I had different parents^), which resulted in Cronbach’s alphas of respectively 0.74, 0.76. Cronbach’s alpha of the alienation scale was 0.62. Higher scores indicated more attachment security. Cognitive Distortions The ‘How I Think Questionnaire’ (HIT, Barriga and Gibbs 1996) was used to assess cognitive distortions of adolescents. The HIT is based upon Gibbs and colleagues’ four-category typology of self-serving cognitive distortions: self-centered attitude; blaming others; minimizing-mislabeling (consequences of) behavior; and assuming the worst (Barriga et al. 2000; Gibbs et al. 1995, 2001). For the present study we used the Dutch validated version of the HIT (Nas et al. 2005). The HIT is composed of 54 items, 39 represent the four types of self-serving cognitive distortions, 8 items are used to screen for anomalous responses, and 7 items are positive filler items. Participants responded on a 6-point scale ranging from agree strongly to disagree strongly. Examples of statements for each subscale are: BIf someone is careless enough to lose a wallet, they deserve to have it stolen (self-centered); People force me to lie when they ask too many questions (blaming others); Everybody breaks the law, it’s not a big deal (minimizing); You should hurt people first, before they hurt you (assuming the worst).^ Scores were averaged across items. In the present study, a total mean score of the four types of selfserving cognitive distortions items was used (39 items). Previous research documented good test-retest reliability for the HIT as well as evidence for construct validity (as described in Barriga et al. 2008). Cronbach’s alpha in the

J Abnorm Child Psychol

present study was found to be 0.91. Higher scores indicated more cognitive distortions. Self-Esteem Feelings of worth and satisfaction with self were measured by using the ‘Competentie Belevingsschaal voor Adolescenten’ (CBSA; Treffers et al. 2002). This questionnaire is a Dutch version of the five-item global self-worth subscale from the ‘Self-Perception Profile for Adolescents’ (SPPA, Harter 1988). Adolescents first chose which of two descriptions described them better (e.g., BSome youngsters are often disappointed in themselves^; BOther youngsters are almost never disappointed in themselves^), then they reported whether that description was a little true or totally true for them. Scores were averaged across items. Higher scores indicated a greater sense of self-worth. The internal consistency of the scale of global self-worth was found to be good, α=0.80 (Evers et al. 2007). The present study’s reliability analysis resulted in a satisfactory Cronbach’s alpha of 0.64. Parental Monitoring Parental knowledge of adolescents’ whereabouts was measured by the ‘Vragenlijst Toezicht Houden’ (VTH; Deković 1996), the Dutch version of the five-item parental monitoring scale used in previous studies (e.g., Brown et al. 1993). Adolescents responded on a 4-point Likert-type scale (1 = nothing, 2 = a little, 3 = a lot, 4 = everything) how much their parents know about who their friends are; how they spent their money; where they were after school; which place they went when they left home; what they did in their leisure time; and what grades they received at school. Scores were averaged across items. Higher scores indicated more parental monitoring. Brown et al. (1993) found an acceptable internal consistency of five-item scale of parental monitoring (α=0.80). The good internal consistency of the scale of parental monitoring was confirmed in present study, α=0.84. Deviant Peer Affiliation Adolescents’ perceptions of deviant peer affiliation were measured by the Dutch version of the ‘Family, Friends and Self Scale’ (‘Delinquentie van Leeftijdgenoten’, Deković 1999; Deković et al. 2004) of Simpson and McBride (1992). Adolescents indicated on 10 items how many of their friends participated in a variety of deviant behaviors (e.g., purposely damage or destroy property) on a scale from 1 (none of my friends) to 5 (almost all of my friends). Scores were averaged across items. Higher scores indicated a higher number of deviant friends. The good internal consistency of the FFS scale was proved by Simpson and McBride (1992). The internal consistency of the scale in present study was found to be excellent, α=0.91. Aggressive Behavior The ‘Buss-Durkee Hostility Inventory’ (BDHI), developed by Buss and Durkee (1957), was used to measure adolescents’ aggression. The BDHI consists of two

subscales ‘Direct Aggression’ (measuring the tendency to express verbal or physical aggression) and ‘Indirect Aggression’ (determining the emotional and cognitive components: hostility, irritability, suspicion, and anger). Results of the present study are based on the two scales of direct and indirect aggression of the Dutch validated 35-item version of the BussDurkee Hostility Inventory (BDHI-D) of Lange et al. (1995). The good internal consistency of the BDHI subscales was proved by previous research (Lange et al. 1995). Cronbach’s alphas of the subscales ‘direct’ and ‘indirect’ aggression in present study were both 0.78 (α=0.85 total scale). Each item was rated as 0 (not true) or 1 (true). Scores were averaged across items. Higher scores indicated higher levels of aggressive behavior. Delinquent Behavior Participation and versatility in delinquency were assessed by the ‘Self-report Delinquency Scale’ (SRD, Van der Laan and Blom 2006). Participants responded on 33 items if they participated in diverse delinquent acts, based on six subscales: property damage, property and theft, violent acts, weapon possession, drugs possession and dealing, and cybercrime. Sum scores of participation in 33 delinquent acts were used for the analyses, with higher scores indicating more delinquent behavior. Cronbach’s alpha was 0.86. Analytic Strategy First, bivariate correlation analyses were conducted between all study variables of the total sample. Next, we tested two separate models for direct and indirect aggression (see Fig. 1) and delinquency (see Fig. 2). The mediating paths of the relation between attachment and (direct/indirect) aggression and between attachment and delinquency were evaluated using structural equation modeling (SEM) techniques. The software package Mplus (Muthén and Muthén 2007) was used to fit the proposed model to the data. Delinquency, cognitive distortions, parental monitoring, deviant peers, self-esteem, and attachment were treated as censored variables. Censored variables are variables with a large fraction of observations at the minimum or maximum value. Many respondents had lower scores on delinquency, deviant peers, cognitive distortions and higher scores on parental monitoring, self-esteem and attachment. The regression coefficients of the censored dependent variables are described as ‘Tobit regression coefficients’ (Tobin 1958). By means of Mplus models with categorical and censored variables with ‘Weighted Least Squares Mean and Variance’ (WLSMV) can be tested. The assessment of SEM models and evaluation of the fit of the models is based on the chi-square (χ2), the corresponding p-value, the comparative fit index (CFI, Bentler 1990), and the root mean squared error of approximation (RMSEA, Steiger and Lind 1980). Good-fitting models show a non-significant

J Abnorm Child Psychol .16 significant not significant Cognitive Distortions

.18 .30

.59

.16

-.41 .06

-.02 Deviant Peers

-.24 .56

Communication

.75 .68

-.08

-.07

-.37

-.18

.41

Attachment

Trust

.82 .69

.56 .31

Direct Aggression

Alienation

.48

Parental Monitoring

Indirect Aggression

-.01

.37

-03 -.12 Self-Esteem

.24

.14

Fig. 1 Structural equation model with standardized parameters estimates: direct and indirect aggression

.16 significant not significant Cognitive Distortions .59

.01

-.40

.45

.06 Deviant Peers -.23 .57

Communication

.63

.76 .83 .68

.08 Attachment

Trust

Delinquency

.30

-.19

.69

.55 Alienation

.48

Parental Monitoring -.11

.37

Self-Esteem

.14

Fig. 2 Structural equation model with standardized parameters estimates: delinquency

J Abnorm Child Psychol

χ2. Values of the RMSEA less than 0.05 are considered to indicate a good fit, with values between 0.05 and 0.08 indicating a fair fit. Values of CFI above 0.90 are generally regarded as evidence for a good fit (Hu and Bentler 1999; Hox and Bechger 1998). Modifications indices (MI’s) were used to guide model specification and improvement of the CFI value (>0.95) or RMSEA value (